Study On Flaws Identification In Ultrasonic Nondestructive Test

Ultrasonic testing is feasible and effective,and it brings an outstanding foreground. As one of the five major nondestructive test(NDT)methods,ultrasonic NDT has been applied in the industry field extensively.An automatically method of flaws identification is proposed in this paper,which combines the advantages of wavelet packet transform(WPT)and empirical mode decomposition(EMD)when adopted to preprocess non-stationary signals,the superiority of the feature evaluation technique based on the average distance within a class and between classes when adopted to select sensitive features,the prominence of the radial basis function(RBF)network when adopted to identify flaws,and the adaptability of D-S evidence theory when adopted to dispose inadequate information.This paper stood on the presentation of the progress and application currently of the relatively technique in ultrasonic NDT.The contributions are summarized as follows:Firstly,overall design scheme of ultrasonic testing system is introduced which included software and hardware.The design of hardware circuit in Multi channel ultrasonic flaw detdction system was emphasis.Additionally ultrasonic probe,couplant and flaw block,and ultrasonic signals were analysed by the software algorithms When PC acquired those.Secondly,the feature extraction and feature evaluation of ultrasonic signals were achieved.That WPT and EMD were combined to apply to ultrasonic signals was ensured.The ultrasonic signals was decomposed by WPT and EMD respectively, and the dimensionless parameters in time domain were extracted from the original signals and each decomposed signal.And then the combined features array1,2and 3 were constructed.The feature evaluation technique which based on the average distance within a class and between classes was introduced.And it was applied to calculate evaluation factors of the combined features.According to the evaluation factors and the identification efficiency of network,the corresponding sensitive features array1,2and3 were selected and input into RBF network.Experiment shows that WPT and EMD could extract features of signals adequately,and it also brings useless features.The method mentioned above has the power of selecting sensitive ones from a large number of features.Accordingly it could overcome the blindness of selecting sensitive features subjectively,and then the scale of network gained reduction,and the accuracy of identification and the efficiency of testing gained enhancement.Thirdly,the primary flaws identification was achieved by RBF network.Basing on the principium and characteristic of RBF network,this paper constructed three RBF networks.The sensitive features array1,2and3 were put into three RBF networks. Accordingly the method mentioned above gave birth to the primary flaws identification in ultrasonic testing.Here it also studied the selecting of parameter correlative,the disposing of input,the training and testing of network,and so on. Experiment shows that the significance of selecting RBF for the network type and constructing several networks were confirmed. RBF networks possess outstanding capability of flaws identification.Finally,the final flaws identification was achieved by D-S evidence theory.In the interest of how to combine RBF network and D-S evidence theory,double fusion model was put forward which was based upon the advantages and shortcomings analyse of both.Experiment shows that this model could reduce the request of network's identification efficiency and also conquer the shortcomings of network more or less.The constructing of D-S evidence theory's mass function is subjective and complicated.Above-mentioned also could escape it.The output is improved when contrasts with the absence of D-S fusing.The final identification accuracy reaches 100% which could meet the need of testing in the industry field.Experiment shows that the method of flaws identification in ultrasonic testing is feasible and effective,and it brings an outstanding foreground.